Year Gradient Dissimilarity Models (nestedness)

This code is Script 4c for Kitchel et al. “Biotic homogenization, the exception and not the rule for marine fish communities” manuscript.

• This project is a collaborative effort to describe changes in taxonomic composition of fish communities around the world–as sampled by bottom trawl surveys.

• Code by Zoë J. Kitchel

SESSION INFO TO DO

library(data.table)
library(vegan)
library(sf)
library(concaveman) #polygon around points
library(betapart) #allows us to partition beta diversity
library(geosphere)
library(ggpubr) #stat_regline_equation
library(nlme)
library(mgcv) #to make gam
library(cowplot)
library(lme4)

#Pull Dissimilarity Means
distances_dissimilarities_allyears.r <- readRDS(here::here("output","dissimilarities", "distances_dissimilarities_allyears.r.rds"))

#make survey and survey unit factors
distances_dissimilarities_allyears.r[,survey:=factor(survey)][,survey_unit:=factor(survey_unit)]

#adjust years
distances_dissimilarities_allyears.r[,year_adj := year-min(year)+1]

#add new variable for year in sequence per region
distances_dissimilarities_allyears.r[,first_year := min(year),.(survey_unit)]
distances_dissimilarities_allyears.r[,last_year := max(year),.(survey_unit)]

#distances_dissimilarities_allyears.r[,year_in_seq := year-first_year+1]

distances_dissimilarities_allyears.r[,years_sampled := last_year-first_year+1]

###Palette for Plotting Palette for plotting all 37 survey units

survey_unit.list <- levels(distances_dissimilarities_allyears.r[,survey_unit])

palette_42 <- c(
  "#5A5156", #AI
  "#DF00DB", #BITS-1
  "#DB8EDA", #BITS-4
  "#F6222E", #CHL
  "#F8A19F", #DFO-NF
  "#16FF32", #DFO-QCS
  "#325A9B", #EBS
  "#3283FE", #EVHOE
  "#FEAF16", #FR-CGFS
  "#fccb6d", #GMEX-Fall
  "#1C8356", #GMEX-Summer
  "#C4451C", #GOA
  "#85660D", #GRL-DE
  "#B0009F", #GSL-N
  "#BF79B8", #GSL-S
  "#1CBE4F", #ICE-GFS
  "#782AB6", #IE-IGFS
  "#90AD1C", #MEDITS
  "#6B003A", #NAM
  "#A75B00", #NEUS-Fall
  "#E3B072", #NEUS-Spring
  "#02E8B6", #NIGFS-1
  "#97E7D5", #NIGFS-4
  "#B00068", #Nor-BTS-3
  "#00B9E3", #NS-IBTS-1
  "#95E2F4", #NS-IBTS-3
  "#B3CE73", #NZ-CHAT
  "#689500", #NZ-ECSI
  "#364d02",#NZ-SUBA
  "#AAF400", #NZ-WCSI
  "#AA0DFE", #PT-IBTS
  "#7f9eb8", #ROCKALL
  "#FA0087", #S-GEORG
  "#DEA0FD", #SCS-Summer
  "#FCEF88", #SEUS-fall
  "#A59405", #SEUS-spring
  "#FCE100", #SEUS-summer
  "#544563", #SWC-IBTS-1
  "#a37fc7", #SWC-IBTS-4
  "#C075A6", #WCANN
  "#BDCDFF", #ZAF-ATL
  "#003EFF"  #ZAF-IND
)

color_link <- data.table(survey_unit = survey_unit.list,hex = palette_42)

Add names for plotting

name_helper <- data.table(Survey_Name_Season = c("Aleutian Islands",
                                    "Baltic Sea Q1",
                                    "Baltic Sea Q4",
                                    "Chile",
                                    "Newfoundland",
                                    "Queen Charlotte Sound",
                                    "Eastern Bering Sea",
                                    "Bay of Biscay",
                                    "English Channel",
                                    "Gulf of Mexico Summer",
                                    "Gulf of Alaska",
                                    "Greenland",
                                    "N Gulf of St. Lawrence",
                                    "S Gulf of St. Lawrence",
                                    "Iceland",
                                    "Irish Sea",
                                    "Mediterranean",
                                    "Namibia",
                                    "NE US Fall",
                                    "NE US Spring",
                                    "N Ireland Q1",
                                    "N Ireland Q4",
                                    "Barents Sea Norway Q3",
                                    "N Sea Q1",
                                    "N Sea Q3",
                                    "Chatham Rise NZ",
                                    "E Coast S Island NZ",
                                    "W Coast S Island NZ",
                                    "Portugal",
                                    "S Georgia",
                                  "Scotian Shelf Summer",
                                  "SE US Fall",
                                  "SE US Spring",
                                  "SE US Summer",
                                  "W Coast US",
                                  "Atlantic Ocean ZA",
                                  "Indian Ocean ZA",
                                   "Rockall Plateau",
                                  "Scotland Shelf Sea Q1",
                                  "Scotland Shelf Sea Q4",
                                  "Falkland Islands",
                                  "Gulf of Mexico Fall",
                                  "Barents Sea Norway Q1",
                                  "Sub-Arctic NZ",
                                  "Scotian Shelf Spring"),
                          survey_unit = c(
                                  "AI",        
                                  "BITS-1",    
                                  "BITS-4",    
                                  "CHL",       
                                  "DFO-NF",    
                                  "DFO-QCS",   
                                  "EBS",       
                                  "EVHOE",     
                                  "FR-CGFS",   
                                  "GMEX-Summer",
                                  "GOA",       
                                  "GRL-DE",    
                                  "GSL-N",     
                                  "GSL-S",     
                                  "ICE-GFS",   
                                  "IE-IGFS",   
                                  "MEDITS",    
                                  "NAM",       
                                  "NEUS-Fall", 
                                  "NEUS-Spring",
                                  "NIGFS-1",   
                                  "NIGFS-4",   
                                  "Nor-BTS-3", 
                                  "NS-IBTS-1", 
                                  "NS-IBTS-3", 
                                  "NZ-CHAT",   
                                  "NZ-ECSI",   
                                  "NZ-WCSI",   
                                  "PT-IBTS",   
                                  "S-GEORG",   
                                  "SCS-SUMMER",
                                  "SEUS-fall", 
                                  "SEUS-spring",
                                  "SEUS-summer",
                                  "WCANN",     
                                  "ZAF-ATL",   
                                  "ZAF-IND",
                                  "ROCKALL",
                                  "SWC-IBTS-1",
                                  "SWC-IBTS-4",
                                  "FALK",
                                  "GMEX-Fall",
                                  "Nor-BTS-1",
                                  "NZ-SUBA",
                                  "SCS-SPRING"
                          ))


color_link <- color_link[name_helper, on = "survey_unit"]

##Make GAMs

Bray Curtis

bray_curtis_gradient_gam <- gam(bray_curtis_dissimilarity_gradient_mean ~ year + s(survey_unit, year, bs = "fs", m = 1),#random smooth
                            data = distances_dissimilarities_allyears.r)

##Make LMERS

Bray These all converged

#running with lme instead of lmer gave same results, but allowed for calculation of p-value
bray_curtis_gradient_lme <- lme(bray_curtis_dissimilarity_gradient_mean ~ year_adj, random = (~1 + year_adj|survey_unit),data = distances_dissimilarities_allyears.r)

#but also run with lmer for confint
bray_curtis_gradient_lmer <- lmer(bray_curtis_dissimilarity_gradient_mean ~ year_adj + (1 + year_adj|survey_unit),data = distances_dissimilarities_allyears.r)

Warning: Model failed to converge with max|grad| = 0.012616 (tol = 0.002, component 1)

summary(bray_curtis_gradient_lme)

Linear mixed-effects model fit by REML
  Data: distances_dissimilarities_allyears.r 

Random effects:
 Formula: ~1 + year_adj | survey_unit
 Structure: General positive-definite, Log-Cholesky parametrization
            StdDev      Corr  
(Intercept) 0.087744906 (Intr)
year_adj    0.001126915 -0.8  
Residual    0.022459795       

Fixed effects:  bray_curtis_dissimilarity_gradient_mean ~ year_adj 

 Correlation: 
         (Intr)
year_adj -0.812

Standardized Within-Group Residuals:
        Min          Q1         Med          Q3         Max 
-3.50705109 -0.58169508 -0.02326812  0.50523697  5.05664006 

Number of Observations: 897
Number of Groups: 42 

anova(bray_curtis_gradient_lme)

bray_curtis_gradient_coefs <- data.table(coef(bray_curtis_gradient_lme))
bray_curtis_gradient_coefs[,survey_unit := rownames(coef(bray_curtis_gradient_lme))][,Year := round(year_adj,5)][,Intercept := round(`(Intercept)`,2)]
View(bray_curtis_gradient_coefs)

bray_curtis_gradient_coefs <- bray_curtis_gradient_coefs[color_link, on = "survey_unit"]

bray_curtis_gradient_coefs.exp <- bray_curtis_gradient_coefs[,.(Survey_Name_Season, Intercept, Year)]

#export this table
fwrite(bray_curtis_gradient_coefs.exp, file = here::here("output","bray_curtis_gradient_coefs.exp.csv"))

Get LMER model as predictions

# need to sort out year in seq versus overall year models
#new data for lmer
lmer_year <- seq(min(distances_dissimilarities_allyears.r[,year]), max(distances_dissimilarities_allyears.r[,year]), by = 0.1)

lmer_year_adj <- seq(min(distances_dissimilarities_allyears.r[,year_adj]), max(distances_dissimilarities_allyears.r[,year_adj]), by = 0.1)

#predict average lmer
lmer_bray_gradient_predictions <- data.table(year = lmer_year, year_adj = lmer_year_adj)

#confidence intervals
bray_curtis_gradient_lmer_confint <- confint(bray_curtis_gradient_lmer)

Computing profile confidence intervals ...

#populate data table of lmer predictions
lmer_bray_gradient_predictions[,bray_curtis_lmer_preds := fixef(bray_curtis_gradient_lmer)[[1]] + fixef(bray_curtis_gradient_lmer)[[2]] * year_adj][,bray_curtis_lmer_preds_lowerCI := bray_curtis_gradient_lmer_confint[5] + bray_curtis_gradient_lmer_confint[6] * year_adj][,bray_curtis_lmer_preds_upperCI := bray_curtis_gradient_lmer_confint[11] + bray_curtis_gradient_lmer_confint[12] * year_adj]

###Move forward with Bray Curtis Gradient (others to supplement)

Coefficients for LMER by survey_unit

#unique survey unit year combos
survey_unit_sampling_years <- unique(distances_dissimilarities_allyears.r[,.(survey_unit, year_adj, year, years_sampled)])

# see group coefficients
model_coefs_reduced <- data.table(transform(as.data.frame(ranef(bray_curtis_gradient_lmer)), lwr = condval - 1.96*condsd, upr = condval + 1.96*condsd))
#https://stackoverflow.com/questions/69805532/extract-the-confidence-intervals-of-lmer-random-effects-plotted-with-dotplotra


#ONLY SLOPES
model_coefs_reduced <- model_coefs_reduced[term == "year_adj",]

model_coefs_reduced[,survey_unit := grp][,year_adj := condval]

#merge with duration of survey
model_coefs_reduced_length <- model_coefs_reduced[survey_unit_sampling_years, on = "survey_unit"]


model_coefs_reduced_length[,years_sampled := as.numeric(years_sampled)][,Directional_Change := ifelse(year_adj > 0, "Differentiation","Homogenization")]

#does it cross zero?
model_coefs_reduced_length[,significant := ifelse(lwr >0 & upr>0,T,ifelse(lwr<0 & upr<0,T,F))]

#significant directional change
model_coefs_reduced_length[,Directional_Change_sig := ifelse(year_adj > 0 & significant == T, "Differentiation",ifelse(year_adj < 0 & significant == T, "Homogenization", "No trend in\ndissimilarity"))]


#min max distances_dissimilarities
min_bray_reduced <- min(distances_dissimilarities_allyears.r[,bray_curtis_dissimilarity_gradient_mean], na.rm = T)
max_bray_reduced <- max(distances_dissimilarities_allyears.r[,bray_curtis_dissimilarity_gradient_mean], na.rm = T)

model_coefs_reduced_length <- model_coefs_reduced_length[color_link, on = "survey_unit"]

#delete any NAs
model_coefs_reduced_length <- na.omit(model_coefs_reduced_length, cols = "significant")

#order table by coefficient
setorder(model_coefs_reduced_length, year_adj)

BC_gradient_model_coefs_reduced_length.unique <- unique(model_coefs_reduced_length[,.(condval,condsd, lwr, upr, survey_unit, year_adj, years_sampled, Directional_Change, hex, Survey_Name_Season, significant, Directional_Change_sig)]) 

#extract color hexes
#year adj coef order
color_year_adj_order <- BC_gradient_model_coefs_reduced_length.unique[,hex]

#alphabetical order
BC_gradient_model_coefs_reduced_length.unique.alpha <- setorder(BC_gradient_model_coefs_reduced_length.unique, Survey_Name_Season)

BC_gradient_model_coefs_reduced_length.unique.alpha[,trend_color := ifelse(Directional_Change_sig == "Homogenization", "#e7ac5b", ifelse(Directional_Change_sig == "Differentiation","#91c874","#cbbfde"))]

color_alpha_order <- BC_gradient_model_coefs_reduced_length.unique.alpha[,hex]
color_alpha_order_bytrend <- BC_gradient_model_coefs_reduced_length.unique.alpha[, trend_color]

saveRDS(BC_gradient_model_coefs_reduced_length.unique, here::here("output","region_stats","BC_gradient_model_coefs_reduced_length.unique.Rds"))

Bar Plot Coefficient LMER

BC_GRADIENT_Dissimilarity_Coef_errorbar_reduced <- ggplot() +
    geom_errorbar(data = model_coefs_reduced_length, aes(x = reorder(Survey_Name_Season, year_adj) , y = year_adj, label = Survey_Name_Season, ymin = lwr, ymax = upr), fill = "grey", width = 0) + #add confidence intervals
  geom_point(data = model_coefs_reduced_length, aes(x = reorder(Survey_Name_Season, year_adj) , y = year_adj, label = Survey_Name_Season, size = years_sampled, fill = Directional_Change_sig, color = Directional_Change_sig), stat = 'identity', shape = 21) +
  scale_fill_manual(values = c("white","black","grey"), name = "Dissimilarity trend", guide="none") +
  scale_color_manual(values = c("black","black","grey"), name = "Dissimilarity trend", guide="none") +
  scale_size_binned(range = c(1,8), name = "Survey period length") +
  geom_hline(yintercept = 0) +
  scale_y_continuous(breaks = seq(-0.005, 0.0075, by = 0.0025), labels = c("-0.005","","0", "", "0.005",  "")) +
  xlab("Survey unit") +
  ylab("β-diversity trend") + #gradient BC dissimilarity trend
  coord_flip() +
  theme_classic() +
  theme(axis.text.y = element_text(colour = color_year_adj_order, face = "bold"), axis.title.y = element_blank(), axis.text.x = element_text(size = 15), axis.title.x = element_text(size = 15), legend.position = c(0.2,0.8), legend.direction = "vertical", legend.background = element_rect(fill = "transparent"))

Warning: Ignoring unknown parameters: `fill`Warning: Ignoring unknown aesthetics: labelWarning: Ignoring unknown aesthetics: labelWarning: Vectorized input to `element_text()` is not officially supported.
ℹ Results may be unexpected or may change in future versions of ggplot2.

#pull legend for homogenization
directional_change_legend_plot <- BC_GRADIENT_Dissimilarity_Coef_errorbar_reduced + 
  scale_fill_manual(values = c("white","black","grey"), name = "Dissimilarity trend") +
  scale_color_manual(values = c("black","black","grey"), name = "Dissimilarity trend") +
  scale_size_binned(range = c(1,8), name = "Survey period length", guide = "none") +
  theme(legend.position = "right", legend.background = element_rect(fill= "transparent")) +
  guides(colour = guide_legend(override.aes = list(size=6)))

Scale for fill is already present.
Adding another scale for fill, which will replace the existing scale.Scale for colour is already present.
Adding another scale for colour, which will replace the existing scale.Scale for size is already present.
Adding another scale for size, which will replace the existing scale.

BC_GRADIENT_Dissimilarity_Coef_errorbar_reduced

#ALT grey scale
BC_GRADIENT_Dissimilarity_Coef_errorbar_reduced_greyscale <- ggplot() +
    geom_errorbar(data = model_coefs_reduced_length, aes(x = reorder(Survey_Name_Season, year_adj) , y = year_adj, label = Survey_Name_Season, ymin = lwr, ymax = upr), fill = "grey", width = 0) + #add confidence intervals
  geom_point(data = model_coefs_reduced_length, aes(x = reorder(Survey_Name_Season, year_adj) , y = year_adj, label = Survey_Name_Season, size = years_sampled, fill = Directional_Change_sig, color = Directional_Change_sig), stat = 'identity', shape = 21) +
  scale_fill_manual(values = c("white","black","grey"), name = "Dissimilarity trend", guide="none") +
  scale_color_manual(values = c("black","black","grey"), name = "Dissimilarity trend", guide="none") +
  scale_size_binned(range = c(1,8), name = "Survey period length") +
  geom_hline(yintercept = 0) +
  scale_y_continuous(breaks = seq(-0.005, 0.0075, by = 0.0025), labels = c("-0.005","","0", "", "0.005",  "")) +
  xlab("Survey unit") +
  ylab("β-diversity trend") + #gradient BC dissimilarity trend
  coord_flip() +
  theme_classic() +
  theme(axis.text.y = element_text(face = "bold"), axis.title.y = element_blank(), axis.text.x = element_text(size = 15), axis.title.x = element_text(size = 15), legend.position = c(0.25,0.8), legend.direction = "vertical")

Warning: Ignoring unknown parameters: `fill`Warning: Ignoring unknown aesthetics: labelWarning: Ignoring unknown aesthetics: label

Alternatively, we color this plot by trend experienced

BC_gradient_Dissimilarity_Coef_errorbar_reduced_colorbytrend <- ggplot() +
    geom_errorbar(data = model_coefs_reduced_length, aes(x = reorder(Survey_Name_Season, year_adj) , y = year_adj, label = Survey_Name_Season, ymin = lwr, ymax = upr), fill = "grey", width = 0) + #add confidence intervals
  geom_point(data = model_coefs_reduced_length, aes(x = reorder(Survey_Name_Season, year_adj) , y = year_adj, label = Survey_Name_Season, size = years_sampled, color = Directional_Change_sig), stat = 'identity') +
  scale_color_manual(values = c("#73BA4D","#E0962C","#cbbfde"), name = "Dissimilarity trend", guide="none") +
  scale_size_binned(range = c(1,8), name = "Survey period length\n(years)") +
  geom_hline(yintercept = 0) +
#  scale_y_continuous(breaks = seq(-0.005, 0.0075, by = 0.0025), labels = c("-0.005","","0", "", "0.005",  "")) +
  xlab("Survey unit") +
  ylab("β-diversity trend") + #gradient BC dissimilarity trend
  coord_flip() +
  theme_classic() +
  theme(axis.text.y = element_text(face = "bold"), axis.title.y = element_blank(), axis.text.x = element_text(size = 15), axis.title.x = element_text(size = 15), legend.position = c(0.25,0.8), legend.direction = "vertical", legend.text = element_text(size = 14), legend.title = element_text(size = 15), legend.background = element_rect(fill = "transparent"))

Warning: Ignoring unknown parameters: `fill`Warning: Ignoring unknown aesthetics: labelWarning: Ignoring unknown aesthetics: label

Wavy Line Plot for GAMs

Generate predicted values

#add colors and names to full dissimilarity data table
distances_dissimilarities_allyears.r <- distances_dissimilarities_allyears.r[color_link, on = "survey_unit"]

#generate new data to smooth lines (need year and season survey combinations)
year_survey_unit_expand.dt <- data.table(survey_unit = as.character(NULL), year = as.numeric(NULL), year_adj = as.numeric(NULL ))

for (i in 1:length(survey_unit.list)) {
  #generate year vectors
  survey_unit_years <- unique(distances_dissimilarities_allyears.r[survey_unit == survey_unit.list[i],.(survey_unit, year, year_adj)])
  
  years <- seq(min(survey_unit_years[,year]), max(survey_unit_years[,year]), by = 0.1)
  
  year_adjust <- seq(min(survey_unit_years[,year_adj]), max(survey_unit_years[,year_adj]), by = 0.1)
  
  year_survey_unit_expand.dt_addition <- data.table(survey_unit = survey_unit.list[i], year = years, year_adj = year_adjust)
  
  year_survey_unit_expand.dt <- rbind(year_survey_unit_expand.dt, year_survey_unit_expand.dt_addition)
}

#add colors and names to full year and survey unit combination table
year_survey_unit_expand.dt <- year_survey_unit_expand.dt[color_link, on = "survey_unit"]

Get model as predictions

#for plotting, get model as predictions
bray_curtis_gradient_gam_predictions <- predict(bray_curtis_gradient_gam, se.fit = TRUE, newdata = year_survey_unit_expand.dt)

#merge into table
year_survey_unit_expand.dt[,bray_glm_mod_fit := bray_curtis_gradient_gam_predictions$fit][,bray_glm_mod_fit_SE := bray_curtis_gradient_gam_predictions$se.fit]

Produce Plot of GAM and mean LMER line

points_wavylines_bray_gradient_year_reduced_gam_nolmer <- ggplot() +
 # geom_ribbon(data = lmer_bray_gradient_predictions, aes(x = year, ymin = bray_curtis_lmer_preds_lowerCI, ymax = bray_curtis_lmer_preds_upperCI), fill = "grey", alpha = 0.2) +
  geom_point(data = na.omit(distances_dissimilarities_allyears.r,cols = "year_adj"),
             aes(x = year,
                 y = bray_curtis_dissimilarity_gradient_mean,
                 color = Survey_Name_Season), alpha = 0.5, size = 1) +
    geom_line(data = na.omit(year_survey_unit_expand.dt,cols = "year_adj"),
             aes(x = year,
                 y = bray_glm_mod_fit,
                 color = Survey_Name_Season)) +
  geom_ribbon(data = na.omit(year_survey_unit_expand.dt,cols = "year_adj"), aes(x = year, ymin=bray_glm_mod_fit-bray_glm_mod_fit_SE, ymax=bray_glm_mod_fit+bray_glm_mod_fit_SE, fill =  Survey_Name_Season), alpha=0.1) + #add standard error
 # geom_line(data = lmer_bray_gradient_predictions, aes(x = year, y = bray_curtis_lmer_preds), color = "black") +
    scale_color_manual(values =  color_alpha_order, name = "Survey Unit") +
  scale_fill_manual(values =  color_alpha_order, guide = "none") +
  theme_classic() +
  lims(x = c(min(distances_dissimilarities_allyears.r[,year]),max(distances_dissimilarities_allyears.r[,year])),
       y = c(0,0.5)) +
  xlab("Year") +
ylab("Gradient BC dissimilarity") +
  theme(legend.position = "null")

points_wavylines_bray_gradient_year_reduced_gam_nolmer

ggsave(points_wavylines_bray_gradient_year_reduced_gam_nolmer, path = here::here("figures"), filename ="points_wavylines_bray_gradient_year_reduced_gam_nolmer.jpg", height = 5, width = 5, unit = "in")

#with lmer

points_wavylines_bray_gradient_year_reduced_gam <- ggplot() +
  geom_ribbon(data = lmer_bray_gradient_predictions, aes(x = year, ymin = bray_curtis_lmer_preds_lowerCI, ymax = bray_curtis_lmer_preds_upperCI), fill = "grey", alpha = 0.3) +
  geom_point(data = na.omit(distances_dissimilarities_allyears.r, cols = "year_adj"),
             aes(x = year,
                 y = bray_curtis_dissimilarity_gradient_mean,
                 fill = Survey_Name_Season), alpha = 0.4, size = 1, shape = 21, color = "white") +
    geom_line(data = na.omit(year_survey_unit_expand.dt, cols = "year_adj"),
             aes(x = year,
                 y = bray_glm_mod_fit,
                 color = Survey_Name_Season), alpha = 0.6) +
  geom_ribbon(data = na.omit(year_survey_unit_expand.dt, cols = "year_adj"), aes(x = year, ymin=bray_glm_mod_fit-bray_glm_mod_fit_SE, ymax=bray_glm_mod_fit+bray_glm_mod_fit_SE, fill =  Survey_Name_Season), alpha=0.1) + #add standard error
  geom_line(data = lmer_bray_gradient_predictions, aes(x = year, y = bray_curtis_lmer_preds), color = "black") +
    scale_color_manual(values =  color_alpha_order, name = "Survey Unit") +
  scale_fill_manual(values =  color_alpha_order, guide = "none") +
  theme_classic() +
  lims(x = c(min(distances_dissimilarities_allyears.r[,year]),max(distances_dissimilarities_allyears.r[,year])),
       y = c(0,0.5)) +
  xlab("Year") +
ylab("β-diversity") +
  theme(legend.position = "null", axis.text = element_text(size = 15), axis.title = element_text(size = 15))

points_wavylines_bray_gradient_year_reduced_gam

ggsave(points_wavylines_bray_gradient_year_reduced_gam, path = here::here("figures"), filename ="points_wavylines_bray_gradient_year_reduced_gam.jpg", height = 6, width = 6, unit = "in")

#ALT
#plot all, but same color scheme (grey)
points_wavylines_bray_gradient_year_reduced_gam_greyscale <- ggplot() +
  geom_ribbon(data = lmer_bray_gradient_predictions, aes(x = year, ymin = bray_curtis_lmer_preds_lowerCI, ymax = bray_curtis_lmer_preds_upperCI), fill = "grey", alpha = 0.3) +
  geom_point(data = distances_dissimilarities_allyears.r,
             aes(x = year,
                 y = bray_curtis_dissimilarity_gradient_mean,
                 fill = Survey_Name_Season), alpha = 0.4, size = 1, shape = 21, color = "white") +
    geom_line(data = year_survey_unit_expand.dt,
             aes(x = year,
                 y = bray_glm_mod_fit,
                 color = Survey_Name_Season), alpha = 0.6) +
  geom_ribbon(data = year_survey_unit_expand.dt, aes(x = year, ymin=bray_glm_mod_fit-bray_glm_mod_fit_SE, ymax=bray_glm_mod_fit+bray_glm_mod_fit_SE, fill =  Survey_Name_Season), alpha=0.1) + #add standard error
  geom_line(data = lmer_bray_gradient_predictions, aes(x = year, y = bray_curtis_lmer_preds), color = "black") +
    scale_color_manual(values =  rep("black", times = length(unique(distances_dissimilarities_allyears.r$Survey_Name_Season))), name = "Survey Unit") +
  scale_fill_manual(values =  rep("black", times = length(unique(distances_dissimilarities_allyears.r$Survey_Name_Season))), guide = "none") +
  theme_classic() +
  lims(x = c(min(distances_dissimilarities_allyears.r[,year]),max(distances_dissimilarities_allyears.r[,year])),
       y = c(0,0.5)) +
  xlab("Year") +
ylab("β-diversity") +
  theme(legend.position = "null", axis.text = element_text(size = 15), axis.title = element_text(size = 15))

points_wavylines_bray_gradient_year_reduced_gam_greyscale

ggsave(points_wavylines_bray_gradient_year_reduced_gam_greyscale, path = here::here("figures"), filename ="points_wavylines_bray_gradient_year_reduced_gam_greyscale.jpg", height = 6, width = 6, unit = "in")

Alternative, color by trend

points_wavylines_bray_gradient_year_reduced_gam_colorbytrend <- ggplot() +
  geom_ribbon(data = lmer_bray_gradient_predictions, aes(x = year, ymin = bray_curtis_lmer_preds_lowerCI, ymax = bray_curtis_lmer_preds_upperCI), fill = "grey", alpha = 0.3) +
  geom_point(data = na.omit(distances_dissimilarities_allyears.r, cols = "year_adj"),
             aes(x = year,
                 y = bray_curtis_dissimilarity_gradient_mean,
                 fill = Survey_Name_Season), alpha = 0.4, size = 1, shape = 21, color = "white") +
    geom_line(data = na.omit(year_survey_unit_expand.dt, cols = "year_adj"),
             aes(x = year,
                 y = bray_glm_mod_fit,
                 color = Survey_Name_Season), alpha = 0.6) +
  geom_ribbon(data = na.omit(year_survey_unit_expand.dt, cols = "year_adj"), aes(x = year, ymin=bray_glm_mod_fit-bray_glm_mod_fit_SE, ymax=bray_glm_mod_fit+bray_glm_mod_fit_SE, fill =  Survey_Name_Season), alpha=0.1) + #add standard error
  geom_line(data = lmer_bray_gradient_predictions, aes(x = year, y = bray_curtis_lmer_preds), color = "black") +
    scale_color_manual(values =  color_alpha_order_bytrend, name = "Survey Unit") +
  scale_fill_manual(values =  color_alpha_order_bytrend, guide = "none") +
  theme_classic() +
  lims(x = c(min(distances_dissimilarities_allyears.r[,year]),max(distances_dissimilarities_allyears.r[,year]))) +
  xlab("Year") +
ylab("β-diversity") +
  theme(legend.position = "null", axis.text = element_text(size = 15), axis.title = element_text(size = 15))

points_wavylines_bray_gradient_year_reduced_gam_colorbytrend

ggsave(points_wavylines_bray_gradient_year_reduced_gam_colorbytrend, path = here::here("figures"), filename ="points_wavylines_bray_gradient_year_reduced_gam_colorbytrend.jpg", height = 6, width = 6, unit = "in")

#plot each independently for supplement
#all survey names = 
all_survey_names <- sort(unique(distances_dissimilarities_allyears.r$Survey_Name_Season))
#list of plots
points_wavylines_bray_gradient_year_reduced_gam_individual <- list()
for (i in 1:length(all_survey_names)) {
points_wavylines_bray_gradient_year_reduced_gam_individual[[i]] <- ggplot() +
  geom_point(data = distances_dissimilarities_allyears.r[Survey_Name_Season == all_survey_names[i]],
             aes(x = year,
                 y = bray_curtis_dissimilarity_gradient_mean), alpha = 0.4, color = "black") +
    geom_line(data = year_survey_unit_expand.dt[Survey_Name_Season == all_survey_names[i]],
             aes(x = year,
                 y = bray_glm_mod_fit), alpha = 0.6) +
  geom_ribbon(data = year_survey_unit_expand.dt[Survey_Name_Season == all_survey_names[i]], aes(x = year, ymin=bray_glm_mod_fit-bray_glm_mod_fit_SE, ymax=bray_glm_mod_fit+bray_glm_mod_fit_SE), alpha=0.1) + #add standard error
  theme_classic() +
#  lims(x = c(min(distances_dissimilarities_allyears.r[Survey_Name_Season == all_survey_names[i],year]),max(distances_dissimilarities_allyears.r[Survey_Name_Season == all_survey_names[i],year])),
#       y = c(0,0.5)) +
  xlab("Year") +
ylab("beta-diversity") +
  facet_wrap(~Survey_Name_Season, ncol = 5) +
  theme(legend.position = "null", axis.text = element_text(size = 15), axis.title = element_text(size = 15))

print(points_wavylines_bray_gradient_year_reduced_gam_individual[[i]])

}

saveRDS(points_wavylines_bray_gradient_year_reduced_gam_individual, here::here("figures","points_wavylines_bray_gradient_year_reduced_gam_individual.Rds"))

#print to pdf
library(gridExtra)

ggsave(
   filename = here::here("figures","points_wavylines_bray_gradient_year_reduced_gam_individual.pdf"), 
   plot = marrangeGrob(points_wavylines_bray_gradient_year_reduced_gam_individual, nrow=1, ncol=1), 
   width = 8.5, height = 11
)

Warning: Removed 1 rows containing missing values (`geom_point()`).Warning: Removed 1 row containing missing values (`geom_line()`).Warning: no non-missing arguments to max; returning -InfWarning: Removed 1 rows containing missing values (`geom_point()`).Warning: Removed 1 row containing missing values (`geom_line()`).Warning: no non-missing arguments to max; returning -InfWarning: Removed 1 rows containing missing values (`geom_point()`).Warning: Removed 1 row containing missing values (`geom_line()`).Warning: no non-missing arguments to max; returning -Inf

Merge BC versus Year plot with GAMS and Region vs. coefficient plot for LMERs

BC_GRADIENT_GAM_LMER_merge_legend <- ggdraw(xlim = c(0, 40.5), ylim = c(0, 21)) +
    draw_plot(points_wavylines_bray_gradient_year_reduced_gam,
                                         x = 1, y = 1, width = 20, height = 20) +
    draw_plot(BC_GRADIENT_Dissimilarity_Coef_errorbar_reduced +
        theme(legend.key.size = unit(0.5, 'cm'), #change legend key size
        legend.title = element_text(size=16), #change legend title font size
        legend.text = element_text(size=14)), #change legend text font size),
                                         x = 20, y = 1, width =19, height = 20) +
    draw_plot(get_legend(directional_change_legend_plot + 
      theme(legend.key.size = unit(0.5, 'cm'), #change legend key size
        legend.title = element_text(size=15), #change legend title font size
        legend.text = element_text(size=13))), #change legend text font size)
                                       x = 27, y = 12, width = 3, height = 2) +
  geom_text(aes(x = 2, y = 20.7), label = ("a."), size =8, fontface = "bold") +
  geom_text(aes(x =20, y = 20.7), label = ("b."), size =8, fontface = "bold")


ggsave(BC_GRADIENT_GAM_LMER_merge_legend, path = here::here("figures"), filename = "BC_GRADIENT_GAM_LMER_merge_legend.png", height = 8, width = 14, units = "in")

#ALT GREY SCALE
BC_GRADIENT_GAM_LMER_merge_legend_greyscale <- ggdraw(xlim = c(0, 40.5), ylim = c(0, 21)) +
    draw_plot(points_wavylines_bray_gradient_year_reduced_gam_greyscale,
                                         x = 1, y = 1, width = 20, height = 20) +
    draw_plot(BC_GRADIENT_Dissimilarity_Coef_errorbar_reduced_greyscale +
        theme(legend.key.size = unit(0.5, 'cm'), #change legend key size
        legend.title = element_text(size=16), #change legend title font size
        legend.text = element_text(size=14)), #change legend text font size),
                                         x = 20, y = 1, width = 19, height = 20) +
    draw_plot(get_legend(directional_change_legend_plot + 
      theme(legend.key.size = unit(0.5, 'cm'), #change legend key size
        legend.title = element_text(size=15), #change legend title font size
        legend.text = element_text(size=13))), #change legend text font size)
                                x = 27, y = 12, width = 3, height = 2) +
  geom_text(aes(x = 2, y = 20.7), label = ("a."), size =8, fontface = "bold") +
  geom_text(aes(x =20, y = 20.7), label = ("b."), size =8, fontface = "bold")

ggsave(BC_GRADIENT_GAM_LMER_merge_legend_greyscale, path = here::here("figures"), filename = "BC_GRADIENT_GAM_LMER_merge_legend_greyscale.png", height = 8, width = 14, units = "in")

#ALT COLOR BY TREND
BC_gradient_GAM_LMER_merge_legend_colorbytrend <- ggdraw(xlim = c(0, 40.5), ylim = c(0, 21)) +
    draw_plot(points_wavylines_bray_gradient_year_reduced_gam_colorbytrend,
                                         x = 1, y = 1, width = 20, height = 20) +
    draw_plot(BC_gradient_Dissimilarity_Coef_errorbar_reduced_colorbytrend +
        theme(legend.key.size = unit(0.5, 'cm'), #change legend key size
       # legend.title = element_text(size=16), #change legend title font size
       # legend.text = element_text(size=14)
       ), #change legend text font size),
                                         x = 20, y = 1, width = 19, height = 20) +
    draw_plot(get_legend(directional_change_legend_plot_colorbytrend + 
      theme(legend.key.size = unit(0.5, 'cm'), #change legend key size
        legend.title = element_text(size=16), #change legend title font size
        legend.text = element_text(size=15))), #change legend text font size)
                                x = 30, y = 6, width = 3, height = 2) +
  geom_text(aes(x = 2, y = 20.7), label = ("a."), size =8, fontface = "bold") +
  geom_text(aes(x =20, y = 20.7), label = ("b."), size =8, fontface = "bold")

ggsave(BC_gradient_GAM_LMER_merge_legend_colorbytrend, path = here::here("figures"), filename = "BC_gradient_GAM_LMER_merge_legend_colorbytrend.png", height = 8, width = 14, units = "in")